ARTICLE IN PRESS

JOURNAL OF
SOUND AND
VIBRATION
Journal of Sound and Vibration 317 (2008) 273–293
www.elsevier.com/locate/jsvi

Whirl and whip instabilities in rotor-bearing system considering

a nonlinear force model
Helio Fiori de Castroa, Katia Lucchesi Cavalcaa,, Rainer Nordmannb
a
Department of Mechanical Design—Unicamp, Postal Box 6122, 13083-970 Campinas, SP, Brazil
b
Department of Mechatronics in Mechanical Engineering, TU-Darmstadt, Petersenstrasse. 30, D-64287 Darmstadt, Germany
Received 17 April 2007; received in revised form 25 February 2008; accepted 27 February 2008
Handling Editor: P. Davies
Available online 8 April 2008

Abstract

Linear models and synchronous response are generally adequate to describe and analyze rotors supported by
hydrodynamic bearings. Hence, stiffness and damping coefficients can provide a good model for a wide range of situations.
However, in some cases, this approach does not suffice to describe the dynamic behavior of the rotor-bearing system.
Moreover, unstable motion occurs due to precessional orbits in the rotor-bearing system. This instability is called ‘‘oil
whirl’’ or ‘‘oil whip’’. The oil whirl phenomenon occurs when the journal bearings are lightly loaded and the shaft is
whirling at a frequency close to one-half of rotor angular speed. When the angular speed of the rotor reaches
approximately twice the natural frequency (first critical speed), the oil whip phenomenon occurs and remains even if the
rotor angular speed increases. Its frequency and vibration mode correspond to the first critical speed. The main purpose of
this paper is to validate a complete nonlinear solution to simulate the fluid-induced instability during run-up and run-
down. A flexible rotor with a central disk under unbalanced excitation is modeled. A nonlinear hydrodynamic model is
considered for short bearing and laminar flow. The effects of unbalance, journal-bearing parameters and rotor
arrangement (vertical or horizontal) on the instability threshold are verified. The model simulations are compared with
measurements at a real vertical power plant and a horizontal test rig.
r 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Studies of the characteristics of rotor-bearing systems are necessary in the design of rotating systems and
their industrial applications, and for the diagnosis of system malfunctions. Mathematical models have been
developed to represent real machines with considerable confidence. Research has also focused on determining
better models of rotating machinery, such as turbogenerators and multistage pumps, which are horizontal
rotating machines with high load capacity. Childs et al. [1] proposed a nonlinear hydrodynamic force model,
based on journal bearing impedance descriptions, which can be used to analyze stability but not to simulate

Corresponding author.
E-mail addresses: heliofc@fem.unicamp.br (H.F. de Castro), katia@fem.unicamp.br (K.L. Cavalca),
nordmann@mim.tu-darmstadt.de (R. Nordmann).

0022-460X/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsv.2008.02.047
 ARTICLE IN PRESS
274 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

Nomenclature R bearing radius

Tz rotor torque
C bearing radial clearance V auxiliary hydrodynamic force function
[C] damping matrix of the system {W} rotor weight
D bearing diameter x radial dimensionless displacement of the
e unbalanced mass eccentricity journal in the bearing
E Young’s modulus X radial displacement of the journal in the
F auxiliary hydrodynamic force function bearing
{Fh} nonlinear hydrodynamic forces y radial dimensionless displacement of the
{Fu} unbalanced forces journal in the bearing
G auxiliary hydrodynamic force function Y radial displacement of the journal in the
[G] gyroscopic matrix of the system bearing
h dimensionless oil film thickness z axial dimensionless displacement of the
H oil film thickness rotor
Jz polar moment of inertia Z axial displacement of the rotor
[K] stiffness matrix of the system Z̄ coordinate reference
L bearing length a auxiliary hydrodynamic force function
Le shaft element length b proportional damping coefficient related
m unbalanced mass to stiffness
[M] mass matrix of the system m dynamic viscosity of oil
p dimensionless oil pressure distribution n cylindrical coordinate of the bearing
P oil pressure distribution r mass density
{q} generalized coordinates. j rotation angle of the rotor
Q dimensionless unbalanced magnitude

fluid-induced instability. Capone [2,3] developed numerical simulations to study cylindrical hydrodynamic
bearings, obtaining the orbits of the shaft in the bearings based on a nonlinear hydrodynamic force model.
Other researchers have investigated the speed-dependent behavior of bearing characteristics. Tiwari et al. [4]
developed an identification algorithm for the characterization of bearing dynamics using unbalanced response
measurements for multidegree-of-freedom flexible rotor-bearing systems. Sinha et al. [5] proposed an
identification method to estimate the static load on the fluid bearings of a flexible machine from rundown
data. This is an important practical problem, because static load strongly influences the bearing’s dynamic
characteristics.
Linear journal-bearing models are normally used to simulate rotor-bearing systems. However, linear models
are not ideal to represent real rotating systems in certain situations. In such cases, nonlinear models should be
employed to better represent real systems [6–9].
The interaction between the rotating system and the oil film in a hydrodynamic bearing causes unstable
dynamic behavior [10], which is characterized by a subsynchronous forward precessional vibration. These
dynamic behaviors are called oil whirl and oil whip [7,11–14].
These instabilities are a kind of self-exciting vibration of a rotor-bearing system. The oil whirl is a
subsynchronous precessional movement that vibrates at a frequency close to half of the rotation speed. When
the rotation speed reaches twice the first natural frequency, the self-excited vibration frequency remains
constant and close to the first resonance frequency. This new behavior is the oil whip instability. Oil whip
instability, in particular, can be severely harmful to the rotor-bearing system. The journal’s amplitude is
limited by the journal bearing clearance, but the shaft vibration amplitude may become very high, since it
vibrates at the natural frequency [11]. The two fluid-induced instability effects can be associated to the bearing
stiffness and the shaft stiffness. If the journal is close to the bearing center, the bearing stiffness is much lower
than the shaft stiffness. In that case, the system stiffness is controlled by the bearing stiffness and the fluid-
induced instability occurs due to oil whirl [15,16]. On the other hand, when the journal is too close to the
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 275

bearing wall, the bearing stiffness is much higher than the shaft stiffness. In that case, the system stiffness is
similar to the shaft stiffness and the instability vibration is caused by the oil whip, which is related to the
bending mode of the rotor system [7].
To reduce the effects of this particular dynamic behavior, some authors have been studying control methods
to reduce oil instability [17].
Jing et al. [18,19] considered the nonlinear model proposed by Capone [2,3] to analyze the nonlinear
dynamic behavior of bearings, taking into account the oil whip phenomenon, although the angular
acceleration was not considered as it is in the present work.
Owing to the nonlinearities of rotating systems supported by lubricated bearings, some researchers
applied nonlinear techniques to study the dynamic behavior of these systems [20–23]. These techniques were
used to identify fluid-induced instabilities demonstrated by experimental results of horizontal and vertical
rotors [24].
Castro et al. [7] introduced angular acceleration in the equation of motion and analyzed fluid-induced
instability (oil whirl and oil whip). They found that Capone’s model is able to model the rotor’s dynamic
behavior under fluid-induced instabilities in transient motion, thus allowing for stability analyses according to
Refs. [11–14].
To continue the study started by Castro et al. [7], an analysis is made taking into account the dynamic
behavior of a rotating system under the oil whirl and oil whip phenomenon, considering the influence of
unbalance, rotor arrangement (vertical or horizontal rotor) and bearing parameters (L/D ratio, bearing
clearance and oil viscosity) on the instability threshold. In addition, the simulated results are compared with
measurements taken at a real vertical power plant and in a horizontal test rig.

2. Mathematical model

A mathematical model of a rotating system can be divided into two parts: the finite element model of the
shaft and concentrated mass of the disk, and the nonlinear hydrodynamic supporting forces of the cylindrical
journal bearing [2,3], which is obtained by solving the Reynolds equation for short bearings. A short journal
bearing scheme and the cylindrical coordinate reference system are shown in Fig. 1.
Eq. (1) describes the pressure distribution P inside the cylindrical journal bearing, based on the solution of
the Reynolds equation for the laminar flow condition:
   2  
q qp R q qp qh dh
h3 þ h3 ¼ þ2 , (1)
qu qW L qz qz qW dt

Fig. 1. Short journal-bearing scheme and coordinate reference system.

 ARTICLE IN PRESS
276 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

where the dimensionless parameters are

 2
H Z X Y P R
_ p¼
h ¼ ¼ 1  x cosðWÞ  y sinðWÞ; z ¼ ; x ¼ ; y ¼ ; t ¼ jt; _
and p0 ¼ 6mj .
C L C C p0 C
The solution of Eq. (1) considers the axial gradient (qp/qz) due to axial losses of lubricating fluid in a short
journal bearing, neglecting the pressure gradient in the angular direction (qp/qW). The bearing parameters are
radial clearance C, bearing length L, bearing radius R and oil viscosity m. The dimensionless oil film thickness
h, dimensionless time t, rotor angular speed j, _ and dimensionless journal displacement in horizontal x,
vertical y and axial z directions are also considered.
The fluid is considered incompressible and the viscosity is assumed to be constant across the fluid. Half
Sommerfeld solution is applied and the cavitation effect is neglected, considering null pressure for negative
values of pressure distribution. Thus, the complete theoretical solution of the differential equation under these
hypotheses is given by
   
1 L 2 ðx  2yÞ_ sinðWÞ  ðy þ 2xÞ
_ cosðWÞ
pðW; zÞ ¼ ð4z2  1Þ. (2)
2 D ð1  x cosðWÞ  y sinðWÞÞ3
In order to determine the hydrodynamic force Fh generated by the distribution of the oil film pressure on
the shaft, the differential contact area (dA ¼ R dWL dz) is considered in the integration, resulting in Eq. (3):
( )  2  2 
F hx R L _ 2 þ ðy þ 2xÞ
½ðx  2yÞ _ 2 1=2
fF h g ¼ ¼  mo ðRLÞ
F hy C2 D2 ð1  x2  y2 Þ
( )
3xV ðx; y; aÞ  sinðaÞGðx; y; aÞ  2 cosðaÞF ðx; y; aÞ
, (3)
3yV ðx; y; aÞ  cosðaÞGðx; y; aÞ  2 sinðaÞF ðx; y; aÞ

where the functions V, G, F and a are, respectively, given in Eqs. (4)–(7):

2 þ ðy cosðaÞ  x sinðaÞÞGðx; y; aÞ
V ðx; y; aÞ ¼ , (4)
ð1  x2  y2 Þ
Z !
aþp
du p 2 y cosðaÞ  x sinðaÞ
Gðx; y; aÞ ¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tan1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , (5)
a ð1  x cosðuÞ  y sinðuÞÞ 1  x2  y 2 1  x2  y 2 1  x2  y 2

ðx cosðaÞ þ y sinðaÞÞ
F ðx; y; aÞ ¼ , (6)
ð1  x2  y2 Þ
   
1 y þ 2x_ p y þ 2x_ p
a ¼ tan  sign _
 signðy þ 2xÞ. (7)
x  2y_ 2 x  2y_ 2
The horizontal rotor scheme is shown in Figs. 2a and b and its finite element model is illustrated in Fig. 2c.
The differential equation of motion is given by Eq. (8), which takes into account the hydrodynamic force {Fh}
given in Eq. (3), the unbalanced forces {Fu}, the rotor weight {W} and the generalized coordinates of the
flexible rotor {q}T ¼ {x1, y1, y, x3, y3, y x6, y6, y, x9, y9, y x11, y11}, which consider all the degrees of
freedom of the rotating system, including the journal displacement in each bearing. If a vertical rotor is
assumed, the rotor weight is not considered:
€ þ ½C½q
½Mfqg _ þ ½Kfqg ¼ fF u g þ fF h g  fW g. (8)
In order to determine the mass [M] and stiffness [K] matrices, a finite element method is used, considering
the shaft and the concentrated mass of the rotor [25,26].
The shaft damping matrix [C] contains a first part proportional to the stiffness matrix and a second part
_
with the gyroscopic effects (½C ¼ b½K þ j½G). The gyroscopic matrix [G] is also obtained by a finite element
method.
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 277

Fig. 2. Horizontal rotor: (a) physical model; (b) physical model—cross-section; (c) finite element model.

Table 1
Material properties for simulation

Material property Value

Young’s modulus, E 2.067  1011 Pa

Density, r 7800 kg/m3
Proportional coefficient related to stiffness, b 25  105

The system excitation force is due to the unbalanced mass m and its eccentricity e, expressed in Eq. (9):
( )
_ 2 cos j
€ sin j þ j
j
fF u g ¼ me . (9)
€ sin j þ jj_ 2 cos j
j
To consider the angular acceleration j€ of the rotating system, Eq. (10) correlates the drive-motor input
torque TZ with the rotation of the system [27], where Jz is the polar moment of inertia. This equation is
obtained by Lagrange’s equation and considers the coupling between torsional and lateral vibrations:
J z j€ ffi T z þ meðsin jx€  cos jy€ þ j
_ x_ cos j þ j_ y_ sin jÞ. (10)
However, this coupling is considerably weak and can be neglected, as indicated by Childs [14]. Eq. (10) is
therefore reduced to Eq. (11):
J z j€ ffi T z . (11)
The solution of the equation of motion is obtained using numerical methods. The Newmark integration
method was chosen here because it is a robust algorithm to solve nonlinear equations in the time domain.

3. Simulation results

3.1. Simulation data

Considering a constant angular acceleration, the length of the four elements between the bearings is
Le ¼ 0.1450 m, the shaft diameter is 0.012 m and the centered disk diameter and length are 0.095 and 0.043 m,
respectively. Simulations were done varying the unbalance and the journal-bearing parameters, as well as the
rotor position (horizontal or vertical). Table 1 lists the material properties considered in the model. A low
 ARTICLE IN PRESS
278 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

proportional structural damping coefficient b was assumed in order to highlight the effects studied in this
work. In all the simulations, the system’s dynamic rundown rather than runup behavior is analyzed, since the
latter may be associated with other phenomena such as the inertial effect of the oil whip.
The dynamic behavior of a rotor supported by journal bearings is important in the design stage. In
predictive maintenance, one can also use the prediction of the dynamic behavior to diagnose failure. The
complete model of the system consists of a linear finite element model for the rotor and a nonlinear model for
the bearings. Before beginning the numerical simulations of the system parameters, the congruence of the
results of the theoretical model and the experimental test rig is ascertained. The influence of design variables
such as unbalance and bearing parameters can be then simulated. The influence of a vertical or horizontal
rotor was also taken into account.

3.2. Horizontal and vertical rotor

First a comparison is made between the horizontal and vertical rotor. Table 2 presents the properties of the
unbalanced moment (me) and the journal bearing. Fig. 3 shows the displacement of the rotor in the bearing for
the horizontal and vertical case in the x and y directions. Waterfall plots must be drawn with the rotor under
acceleration in order to observe the dynamic behavior in the frequency domain. This type of diagram

Table 2
Parameters for the comparison of horizontal and vertical behavior

Unbalance moment me 2  105 kg m

Bearing clearance, C 90 mm
Bearing diameter, D 0.031 m
Bearing length, L 0.010 m
Oil viscosity, m 0.04 Pa s

Fig. 3. (a) Displacement x—horizontal rotor, (b) displacement y—horizontal rotor, (c) displacement x—vertical rotor; (d) displacement
y—vertical rotor.
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 279

correlates the rotating speed, signal magnitude and frequency spectrum, indicating if the precessional
vibration is dominated by a synchronous or asynchronous vibration or even by a combination of them.
Considering the displacements in the bearing, waterfall plots are generated for each case, as shown in Fig. 4a
for a horizontal rotor and in Fig. 4b for a vertical rotor.
The resonance frequency of the rotor is close to 22 Hz, as illustrated in Fig. 3, which indicates that
precessional vibration increases in both cases when the rotation speed approaches 22 Hz. Fig. 4 shows a
similar result because both cases (horizontal and vertical) present a 1.0  increase in vibration amplitude at a
frequency of 22 Hz.
In the vertical rotor, the shaft vibrates around the geometric center of the bearing. However, at low
rotational speeds, this equilibrium point in the horizontal rotor is not in the bearing’s geometric center, as
indicated in Figs. 3a and b. The amplitude at zero Hz in the waterfall plot of the horizontal rotor (Fig. 4a)
represents the equilibrium position of the shaft. This position is far from the bearing center at lower rotation
speeds, but close to the center in the oil whip region (above 45 Hz).
To link the waterfall diagrams (Fig. 4) to the dynamic behavior of the rotor-bearing system, some rotation
speed intervals are analyzed. In the horizontal rotor, the first region occurs before the resonance frequency,
below 20 Hz. The displacements shift towards the center of the bearing center as the rotation speed increases,
as illustrated in Figs. 3a and b. The second region occurs between 20 and 25 Hz, when the synchronous
vibration assumes higher values and the rotor crosses over the resonance. The vibration in the third region
located above the resonance (between 25 and 45 Hz) is of lower amplitude and the oil whirl precessional
vibration is not evident here. One of the reasons for this behavior is the influence of the system’s weight, which
leads to a position of equilibrium and, hence, to stable motion. The fourth region, characterized by the highest
vibration amplitude caused by the oil whip, occurs above 45 Hz. Here, the shaft equilibrium position is at the
bearing’s geometric center.
The vertical rotor displacements generally occur around the center of the bearing, as indicated in Figs. 3c
and d. Similarly to the horizontal case, four regions are also visible here: the region before the resonance
frequency at 20 Hz, the resonance region between 20 and 25 Hz, the oil whirl region from 30 to 45 Hz, and
the fourth region, above 45 Hz, which suffers the oil whip effect and where amplitudes increase significantly
(Fig. 4b).

Fig. 4. Waterfall diagrams for me ¼ 2  105 kg m: (a) horizontal rotor; (b) vertical rotor.
 ARTICLE IN PRESS
280 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

The synchronous vibration decreases above resonance in the vertical rotor, at which point the
subsynchronous vibration enters into action. From 30 to 45 Hz (Fig. 4b), the vibration frequency is close
to one half of rotation speed, which is characterized by oil whirl. Above 45 Hz (or twice the resonance
frequency), the vibration frequency remains constant and equal to the resonance frequency (22 Hz). This is oil
whip behavior. In the horizontal rotor, oil whirl is not noticeable in the same frequency range. Instead, oil
whip appears at close to double the first resonance frequency (45 Hz) because the system’s dynamic behavior is
stable up to this frequency due to the rotor weight effect (Fig. 4a).
The main reason for the distinct behaviors of these two assemblies (horizontal and vertical) is the journal’s
position in the bearing. Because the journal usually closer to the bearing center in the vertical rotor, the oil
whirl effect is more evident in the vertical assembly. In contrast, in the horizontal rotor, the journal vibrates
around the journal’s position of equilibrium in the bearing (shaft locus), which is not as close to the center of
the bearing as in the vertical rotor (see Figs. 3a and b).

3.3. Unbalance influence

The influence of unbalance was investigated in the vertical rotor because the instability is more evident in
this assembly. The journal-bearing parameters listed in Table 2 were taken into account in the numerical
simulations.
Fig. 5 illustrates the progression of the deflection in the journal bearing as the rotation speed increases.
Different unbalanced moments are considered for each curve. The analysis is aided by waterfall plots for some
conditions to better illustrate this progression. The waterfall plot in Fig. 4b represents the lowest unbalanced
moment (me ¼ 2  105 kg m), while the waterfall plots in Figs. 6a and b show moments of greater unbalance
of 5 and 7.5  105 kg m, respectively.
The subsynchronous motion (0.5  ) lasts longer for lower unbalanced moments, due to the weak forward
synchronous vibration at 1.0  . When the moment of unbalance increases, the amplitude of the synchronous
vibration increases and dominates the precessional motion of the rotor. When the rotation speed exceeds the
first resonance frequency at 22 Hz, the precession becomes subsynchronous. Because the unbalance is higher,
the beginning of subsynchronous movement shifts to higher rotational speeds, occurring even closer to double
the first resonance frequency where the rotor vibrates in its natural frequency, which does not change in the
frequency domain. This effect is called oil whip instability.
To analyze how unbalance affects the nonlinear behavior of the rotor-bearing system, bifurcation diagrams
were drawn for 32 and 35 Hz. These two rotation speeds were chosen because instability begins in this range of
frequencies and the amplitude increases significantly, as indicated in Fig. 5. Figs. 7a and b, respectively, depict

Fig. 5. Nonstationary deflection of the rotor considering several unbalanced values (solid line—me ¼ 2  105 kg m; dotted line—
me ¼ 5.0  105 kg m; dashed line—me ¼ 7.5  105 kg m).
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 281

Fig. 6. Waterfall diagrams: (a) me ¼ 5.0  105 kg m; (b) me ¼ 7.5  105 kg m.

Fig. 7. (a) Bifurcation diagram for unbalance influence: (a) horizontal displacement at 32 Hz; (b) vertical displacement at 32 Hz;
(c) horizontal displacement at 35 Hz; (d) vertical displacement at 35 Hz (C ¼ 90 mm, L/D ¼ 0.5 and m ¼ 0.04 Pa s).

bifurcation diagrams of horizontal and vertical displacement at 32 Hz, while Figs. 7c and d show these
displacements at 35 Hz.
At 32 Hz, the unbalance threshold of instability is 1.5  104 kg m. A period-doubling bifurcation is
encountered above this threshold, which means that there is a transition from a synchronous whirl to a
 ARTICLE IN PRESS
282 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

subsynchronous whirl at approximately half the rotating speed, indicating oil whirl instability [23]. Fig. 8
shows the rotor orbit and Poincaré sections. In the orbit where the unbalance is 1.0  104 kg m, the whirl is
subsynchronous due to the period-doubling motion described by the Poincaré sections. The transition from
period doubling to period motion is clearly visible when the unbalance is 1.5  104 kg m. At an unbalance of
more than 1.5  104 kg m, the whirl is synchronous (period motion).
When the rotation speed is 35 Hz, the unbalanced threshold of instability occurs at 3  104 kg m, indicating
that when the unbalance is higher, the beginning of subsynchronous movement shifts to higher rotational
speeds.
It is therefore clear that the amount of unbalance significantly affects the dynamic behavior of the rotor in
terms of fluid-induced instabilities. The findings reported here are congruent with the experimental results of
Muszynska [11], who demonstrates that higher unbalances produce a wider range of rotations with stable
synchronous vibrations, i.e., the subsynchronous vibration is imperceptible.

3.4. Journal-bearing parameters

Other parameters are checked due to their significant influence on the dynamic behavior of the system with
respect to hydrodynamic bearings: radial clearance, viscosity of the lubricating oil and L/D ratio (length-to-
diameter ratio of the bearing). These parameters are related to the Ocvirk or nondimensional Sommerfeld
numbers [14].
However, an analysis of the system’s response to the changes in these parameters would not suffice to
understand the dynamic behavior of the system, because of the nonlinearities inherent to the rotor-bearing
system. Bifurcation diagrams, orbits and Poincaré sections were therefore drawn to illustrate the influence of
radial clearance C (Figs. 9 and 10), lubricant viscosity m (Figs. 11 and 12) and L/D ratio (Figs. 13 and 14).
Increased radial clearance leads to a period-doubling motion, as indicated in Figs. 9 and 10. The transition
from a synchronous to a subsynchronous whirl begins when the radial clearance is 80 mm. At a radial clearance
of 90 mm, the influence of the half-speed frequency is much higher and at 100 mm the motion is practically
period doubling. The same effect is indicated by the viscosity and L/D ratio. Fig. 11 shows the bifurcation

Fig. 8. Orbit and Poincaré sections at 32 Hz for unbalance of 1  104, 1.5  104, 2  104 and 5  104 kg m.
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 283

Fig. 9. Bifurcation diagram for influence of radial clearance: (a) horizontal displacement; (b) vertical displacement (me ¼ 1.5  104 kg m,
L/D ¼ 0.5 and m ¼ 0.04 Pa s).

Fig. 10. Orbit and Poincaré sections at 32 Hz for radial clearance of 80, 90, 95 and 100 mm.
 ARTICLE IN PRESS
284 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

Fig. 11. Bifurcation diagram for influence of viscosity: (a) horizontal displacement; (b) vertical displacement (C ¼ 90 mm, L/D ¼ 0.5,
me ¼ 1.5  104 kg m).

Fig. 12. Orbit and Poincaré sections at 32 Hz for viscosity of 0.010, 0.025, 0.040 and 0.055 Pa s.
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 285

Fig. 13. Bifurcation diagram for L/D ratio: (a) horizontal displacement; (b) vertical displacement (C ¼ 90 mm, m ¼ 0.04 Pa s,
me ¼ 1.5  104 kg m).

Fig. 14. Orbit and Poincaré sections at 32 Hz for L/D ratio of 0.5, 0.54, 0.58 and 0.625.
 ARTICLE IN PRESS
286 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

diagram and Fig. 12 shows the orbits and Poincaré sections for the variation in viscosity. The motion is
synchronous if the viscosity is 0.01 Pa s, but period-doubling motion starts if the viscosity changes to 0.015,
indicating the beginning of oil whirl instability. Figs. 13 and 14 show these diagrams for the increase in L/D
ratio. At a ratio of 0.5, the unstable motion is at the threshold of instability, showing the transition from
synchronous to subsynchronous motion. Another effect that can be observed when the viscosity and L/D ratio
increase is the decrease in displacement amplitude, see Figs. 11 and 13. A comparison with Eq. (3) reveals that
the higher the viscosity or L/D ratio the higher the hydrodynamic supporting forces. Thus, the stiffness and
damping effect are also greater, reducing the amplitude of the displacement.
Therefore, increasing these bearing parameters leads to instable motion, which cannot be deter-
mined considering only their relation given by Ocvirk or nondimensional Sommerfeld numbers. The
reduction in displacement amplitude due to higher viscosity and L/D ratio is also congruous with these
numbers.

Fig. 15. Schematic of real power plant rotor.

 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 287

Fig. 16. Unbalance effect of oil-whirl instability in a real turbine ( Q ¼ 1.5; Q ¼ 6.3; Q ¼ 15).

4. Comparative analysis with a real vertical power plant

The main purpose of the comparative analysis presented here is to illustrate the similarity of the simulated
results with measurements taken in a real power plant whose dynamic characteristics closely resemble those of
the simulated rotor (Fig. 15). To this end, the behavior of the hydrodynamic cylindrical bearing assembled in
the power plant rotor is compared with the computational simulations obtained by the vertical rotor model
proposed here. The upper hydrodynamic short bearing is assembled under the main rotor, with a radial
clearance of 80 mm. The natural frequency of the rotor is 60 Hz (3600 rev/min), with oil-whip instability at
120 Hz (7200 rev/min).
Fig. 16 shows the influence of the dimensionless unbalance Q (applied unbalance/residual unbalance) on the
turbine stability (oil whirl). It is interesting to note the similarity to the simulated results depicted in Fig. 5,
except for the effect of the greater flexibility in the simulated rotor, which is revealed by the amplitude peaks at
the natural frequency of the rotor (22 Hz). However, one can clearly see that the same unbalance effect occurs
in both systems (real turbine and above 35 Hz for simulated rotor): when the mass unbalance increases, the oil
whirl is delayed in the frequency domain with lower vibration amplitudes. Therefore, the nonlinear
hydrodynamic-bearing model applied here yields very satisfactory results, providing robustness to its
application in critical situations, when rotor vibration causes the bearings to work with very thin oil film
viscosity.

5. Comparative analysis with a horizontal test rig

The experimental setup consisted of two hydrodynamic bearings and a lumped mass assembled in the center
of the shaft (Fig. 17), like the Laval rotor used in the simulation tests. The shaft of the physical system was
made of steel with a 12 mm diameter. The distance between the bearings was 600 mm. The concentrated mass
consisted of a disk with an external diameter of 94.7 mm, a length of 47.0 mm and a mass of 2.34 kg. A pair of
cylindrical hydrodynamic bearings was used to support the shaft, which was made of in bronze, with a radial
clearance of 90 mm and L/D ratio of 0.64. The bearings were lubricated with AWS 32 oil, whose operating
temperature was close to 25 1C. The nominal frequency of the electric motor lay in the range of 1–60 Hz. The
rotor support structure can be considered rigid in these operational frequencies.
The unbalance time response was recorded at the concentrated mass of the rotor. In order to obtain the
curves in the horizontal and vertical directions, the orbit of the rotor mass was monitored by two magnetic
proximity sensors. To monitor the displacement of the shaft inside the bearings, two magnetic proximity
sensors were immersed in the oil film. The measurements were taken within the range of the nominal rotation
speed of the electric motor. Fig. 18 shows the variation of the rotation speed over time in a range of 3–50 Hz.
A fast runup is carried out first, followed by a slower rundown. The experimental responses (horizontal and
vertical directions) are compared with the simulation results. Figs. 19 and 20 show, respectively, the
experimental measurements and the simulation results for bearing 1. Figs. 21 and 22 show these results for
bearing 2, while Figs. 23 and 24 depict the results for the concentrated mass. In each case, the simulated and
 ARTICLE IN PRESS
288 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

Fig. 17. Experimental test rig.

Fig. 18. Variation of rotational speed in experimental data acquisitions.

experimental results are very similar, indicating that the nonlinear journal-bearing model satisfactorily
represents the fluid-induced instability effects.
The experimental results of bearings 1 and 2 show a synchronous frequency close to 45 Hz, which is related
to the bearing housing first mode shape. Neither this nor the second harmonic effects are taken into account in
the numerical simulations.

6. Conclusion

A nonstationary model of a rotor-bearing system is proposed, considering nonlinear hydrodynamic forces.

Vertical and horizontal rotors were simulated. Different moments of unbalance were applied to ascertain their
influence on the rotor’s dynamic behavior and instability threshold. The influence of hydrodynamic journal-
bearing parameters such as oil film viscosity, radial clearance and bearing length and diameter ratio were also
analyzed.
The bearing’s model robust evaluation plays a significant role in the identification processes that involves
parallel linear and nonlinear systems. The proposed model is used in numerical simulations of the rotor run up
and run down to analyze the dynamic behavior of the bearings. The journal-bearing nonlinear model [2,3] is
able to simulate oil whirl and oil whip, showing a promising potential to reproduce displacements and
hydrodynamic journal-bearing forces.
The subsynchronous whirl is more evident in the vertical rotor. As shown in Fig. 4b, the oil whirl and oil
whip are well characterized. In the horizontal rotor, the whirl instability is not as evident as in the vertical
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 289

Fig. 19. Experimental results for bearing 1: (a) horizontal displacement; (b) vertical displacement.

Fig. 20. Simulated results for bearing 1: (a) horizontal displacement; (b) vertical displacement.

rotor, but the whip instability clearly appears at close to double the natural frequency (Fig. 4a). The influence
of unbalance was studied in vertical rotors because they highlight the whirl effect. Increasing the moment of
unbalance increases the threshold of instability. At the lowest moment of unbalance, synchronous precessional
vibration was hardly perceptible in the waterfall plot. As the unbalance moment increased, this kind of
 ARTICLE IN PRESS
290 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

Fig. 21. Experimental results for bearing 2: (a) horizontal displacement; (b) vertical displacement.

Fig. 22. Simulated results for bearing 2: (a) horizontal displacement; (b) vertical displacement.

vibration became more evident and the region of stability increased. These diagrams are congruent with
Muszynska [11,12].
Bifurcation diagrams and Poincaré sections were used to analyze the influence of unbalance and journal-
bearing parameters (radial clearance C, viscosity m and ratio L/D) in the nonlinear dynamic behavior of the
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 291

Fig. 23. Experimental results for concentrated mass: (a) horizontal displacement; (b) vertical displacement.

Fig. 24. Simulated results for concentrated mass: (a) horizontal displacement; (b) vertical displacement.

system’s response. The decrease of unbalance and increase of all bearing parameters caused a shift from a
period motion, characterized by synchronous vibration due to unbalance excitation, to a period-doubling
motion, which is characteristic of the half-speed frequency vibration and indicates oil whirl. The orbits and
 ARTICLE IN PRESS
292 H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293

Poincaré sections show the transition from synchronous to subsynchronous vibration at the threshold of
instability.
A comparison was made with a real case study involving a vertical power plant and a horizontal test rig. In
the comparison with the vertical power plant, the effect of the unbalance on the oil whirl instability was quite
similar to the simulations. Experimental measurements of the test rig were carried out at both bearings and the
concentrated mass, showing satisfactory results. This comparison confirmed that the nonlinear hydrodynamic
journal bearing model offers a viable solution for rotor-bearing system simulations when considering fluid-
induced instabilities.

Acknowledgments

The authors gratefully acknowledge the Brazilian and German research funding agencies CAPES/DAAD
PROBRAL program, CNPq and FAPESP for their support of this work.

References

[1] D. Childs, H. Moes, H. Leeuwen, Journal bearing impedance descriptions for rotordynamic applications, Journal of Lubrification
Technology (1977) 198–214.
[2] G. Capone, Orbital motions of rigid symmetric rotor supported on journal bearings, La Meccanica Italiana 199 (1986) 37–46.
[3] G. Capone, Analytical description of fluid-dynamic force field in cylindrical journal bearing, L’Energia Elettrica 3 (1991) 105–110 (in
Italian).
[4] J.K. Tiwari, A.W. Lees, M.I. Friswell, Identification of speed-dependent bearings parameters, Journal of Sound and Vibration 254
(2002) 967–986.
[5] J.K. Sinha, A.W. Lees, M.I. Friswell, Estimating the static load on the fluid bearings of a flexible machine from run-down data,
Mechanical System and Signal Processing 18 (2004) 1349–1368.
[6] H. Diken, Non-linear vibration analysis and subharmonic whirl frequencies of the Jeffcott rotor model, Journal of Sound and
Vibration 243 (2001) 117–125.
[7] H.F. Castro, K.L. Cavalca, R. Nordmann, Rotor-bearing system instabilities considering a non-linear hydrodynamic model, Seventh
IFToMM-Conference on Rotor Dynamics, Vienna, September 2006, pp. 1–10.
[8] K.L. Cavalca, F.G. Dedini, Experimental analysis of tilting-pad journal bearing influence in a vertical rotating system, IFToMM-
International Conference on Rotor Dynamics, Darmstadt, September 1998, pp. 571–582.
[9] E.P. Okabe, K.L. Cavalca, Rotordynamic analysis if system with a non-linear model of tilting pad bearings, Seventh IFToMM-
Conference on Rotor Dynamics, Vienna, September 2006, pp. 1–10.
[10] Robert Gasch, Rainer Nordmann, Herbert Pfützner Rotordynamik, Springer, Berlin, 2002.
[11] A. Muszynska, Whirl and whip—rotor bearing stability problems, Journal of Sound and Vibration 110 (3) (1986) 443–462.
[12] A. Muszynska, Stability of whirl and whip in rotor bearing system, Journal of Sound and Vibration 127 (1) (1988) 49–64.
[13] S. Crandall, From whirl to whip in rotordynamics, Transactions of IFToMM Third International Conference on Rotordynamics, Lyon,
September 1990, pp. 19–26.
[14] Dara Childs, Turbomachinery Rotordynamics, Wiley-Intersciences, New York, 1993.
[15] A. Muszynska, D.E. Bently, Fluid-generated instabilities of rotors, Bently Nevada Corporation ORBIT 10 (1) (1989) 6–14.
[16] D.E. Bently, The death of whirl and whip, Bently Nevada Corporation ORBIT, Fourth Quarter, 2001, pp. 42–46.
[17] B.-H. Rho, K.-W. Kim, A study of the dynamic characteristic of synchronous controlled hydrodynamic journal bearing, Tribology
International 35 (2002) 339–345.
[18] J. Jing, G. Meng, Y. Sun, S. Xia, On the non-linear dynamic of a rotor-bearing system, Journal of Sound and Vibration 274 (2004)
1031–1044.
[19] J. Jing, G. Meng, Y. Sun, S. Xia, On the whipping of a rotor-bearing system by a continuum model, Applied Mathematical Modelling
29 (2004) 461–475.
[20] G. Adiletta, A.R. Guido, C. Rossi, Nonlinear dynamics of a rigid unbalanced rotor in journal bearings—part I: theoretical analysis,
Nonlinear Dynamics 14 (1997) 57–87.
[21] G. Adiletta, A.R. Guido, C. Rossi, Nonlinear dynamics of a rigid unbalanced rotor in journal bearings—part II: experimental
analysis, Nonlinear Dynamics 14 (1997) 157–189.
[22] J.K. Wang, M.M. Khonsari, Bifurcation analysis of a flexible rotor supported by two fluid-film journal bearing, Journal of Tribology
128 (2006) 594–603.
[23] Q. Ding, A.Y.T. Leung, Numerical and experimental investigation on flexible multi bearing rotor dynamics, Journal of Vibration and
Acoustics 127 (2005) 408–415.
[24] M.A.C. Michalski, Y.M.B. Moreno, R.O. Rocha, M. Zindeluk, Identifying oil instability phenomena in rotors using nonlinear
techniques analysis, Proceedings of XII International Symposium on Dynamic Problems of Mechanics, Ilhabela, Brazil, 2007, pp.1–10.
 ARTICLE IN PRESS
H.F. de Castro et al. / Journal of Sound and Vibration 317 (2008) 273–293 293

[25] R.L. Ruhl, J.F. Booker, A finite element model for distributed parameter turborotor systems, Journal of Engineering for Industry 94
(1972) 128–132.
[26] H.D. Nelson, J.M. McVaugh, Dynamics of rotor-bearing systems using finite elements, Journal of Engineering for Industry 98 (1976)
593–600.
[27] K.L. Cavalca, Non-linear analysis of a Laval Rotor passing through the resonance under a Coulomb dry-friction action, Machine
Vibration 2 (2) (1993) 71–79.